Choropleth Maps (Draft)

Choropleth mapping is a cartographic technique that is based on applying a colour or shading scheme to areal geographic data in order to represent the data in a particular way. Because the method of representation is area-based, it is ideal for representing area-specific count data that have been normalised (such as densities, proportions, ratios or rates) (Axis Maps, n.d.a). However, without consideration for how data are classified, choropleth maps can (accidentally or intentionally) misrepresent the data, and distort the story being told by the map (Monmonier, 2005, pp 217 - 220).

This story map looks at various different methods of data classification, and discusses the effects of each on the representation of the data. This is presented through a series of maps that represent the youth population density alongside the distribution of playgrounds in Wellington's central and southern suburbs. Youth population density is calculated by counting the population aged under 15 in each statistical area, and normalising this by the area (in square kilometres) of each statistical unit. This results in a normalised density figure of "number of people aged under 15 per square kilometre". Areas with a higher density of youth population are shown in a darker colour, where those with a lower density are shown in a lighter colour. Playgrounds are shown as red points (Wellington City Council, 2020).

An unclassified choropleth map uses a continuous colour spectrum that stretches from the lowest data value (represented by the lightest colour) through to the highest data value (represented by the darkest colour). Each areal unit is coloured in accordance with where its data value falls on the spectrum.

In this instance, the unclassified map above reveals some unusually high "outlier" densities around the suburb of Newtown. The majority of the data however is at the lower end of the spectrum, and difficult to distinguish individually.

A natural breaks classification uses algorithms to group data into a specified number of classes (five in this case) (Monmonier, 2006, p218). Data is grouped in order to minimise difference within each class, and maximise difference between classes (Axis Maps, n.d.b).

When compared to an unclassified map, broader patterns begin to emerge in a map classified using natural breaks. In the example above, there is a particularly clear distinction between areas with the lowest class of density, and those classes above. At the same time, the "outlier" densities that were apparent in the unclassified map recede and become difficult to distinguish.

The equal interval classification method evenly divides the full spectrum of data into classes of equal size (Axis Maps, n.d.b).

In the map above, most of the data being presented sits around the lower end of the range of densities. Because of this, most of the data is grouped into one class (in the lightest colour), with a small number of higher density outliers being revealed as higher classes. In this instance, the equal interval classification is only useful for identifying high-density outliers, while the remainder of the data is rendered meaningless.

Quantile classifications place an equal number of data points in each class (Axis Maps, n.d.b). In the map above, the data for youth population density is sits mostly at the low end of the range, with a small number of high density outliers. Using the quantile classification method, the data range of the class representing the highest density becomes very large. As a result, the map gives the impression that there is a greater degree of density overall, when compared to other classification methods.

The manual classification method allows the cartographer to specify class sizes and breaks independent of the distribution of the data. Theoretically the cartographer could apply any logic to the classification, or even try and match some of the other classification methods discussed above.

The example above presents a manual classification scheme that groups a large number of data points together in the lowest class (effectively hiding them), revealing "pockets" of medium and higher densities through groupings of higher classes.

Looking at the examples presented above, it is apparent that different choropleth schemes will be more or less effective at communicating certain messages.

The natural breaks classification method looks useful for identifying initial patterns in the distribution of youth population density its relationship to the location of existing playgrounds in Wellington. This might, for example, be the starting point for an investigation into whether or not the overall distribution of existing playgrounds is appropriate. Because the data is subject to high-density outliers, the equal interval and unclassified choropleths could be useful for identifying the location of these outliers for further investigation. And if the Wellington City Council had set criteria for identifying the location of new playgrounds in relation to particular youth population densities, then applying these criteria to a manual break classification would result in a useful choropleth to inform playground planning.

In the end, whether or not a particular classification scheme is more or less appropriate will depend on the purpose of the map.